The U.S. owes $39 trillion. AI-driven infrastructure is the only plausible engine to outrun the debt. A datacenter moratorium doesn't pause progress — it cedes America's position.
The United States owes $39 trillion. The national debt crossed that threshold on March 17, and most Americans didn't notice because the number stopped meaning anything to them years ago. It is, by every serious measure, the single greatest threat to American sovereignty, prosperity, and security — and it is accelerating.
The Congressional Budget Office projects debt will reach 120% of GDP by 2036 and 175% by 2056. Interest payments alone consumed $1 trillion in fiscal year 2025 — more than the federal government spends on Medicare, national defense, or any other line item except Social Security. By 2031, the average interest rate paid on federal debt will exceed the economy's rate of growth. Economists have a clinical term for that: a debt spiral. Once borrowing costs grow faster than GDP, there is no combination of spending cuts or tax increases that makes the math work without extraordinary economic growth.
That is not a political opinion. It is arithmetic.
And yesterday, two members of Congress introduced a bill to kneecap one of the only plausible engines that could generate enough GDP growth to outrun the debt.
The Moratorium
Sen. Bernie Sanders and Rep. Alexandria Ocasio-Cortez introduced legislation this week to impose a moratorium on new datacenter construction until worker, consumer, and environmental safeguards are implemented. The bill will not pass. Both sponsors know that. Sen. John Fetterman responded the same day, calling the proposal “China First.”
But the thinking behind the bill — that America should slow down its digital infrastructure buildout to study the consequences — represents something more dangerous than a symbolic gesture. It represents a fundamental misunderstanding of what is actually keeping the American economy upright.
The Disease
America's fiscal position is worse today than it has been at any point in the nation's history, including the peak of World War II spending. When the dot-com bubble burst, debt stood at 34% of GDP. When the 2008 financial crisis hit, it was 35%. The government had room to borrow, stimulate, and recover.
That room is gone.
The gross national debt has increased by $2.25 trillion in the past year alone — roughly $8 billion per day, $334 million per hour, $5.57 million per minute. Every American household's share is approximately $285,000. The deficit-to-GDP ratio is projected to average 6.1% over the next decade, more than double the 3% threshold that Treasury Secretary Bessent has publicly targeted.
Former Fed Chair Janet Yellen has warned that the U.S. is approaching “fiscal dominance” — the point at which financing the debt begins to constrain the Federal Reserve's ability to fight inflation. When debt was 25% of GDP, Paul Volcker could raise interest rates to crush inflation without the interest payments themselves becoming destabilizing. At 120% of GDP, raising rates makes the deficit worse, which makes the debt worse, which requires more borrowing at higher rates. The brake pedal becomes an accelerator.
The Committee for a Responsible Federal Budget warns the country has never experienced an economic shock this indebted. Their March 2026 report calls for a pre-negotiated emergency fiscal blueprint — a “Break Glass Plan” — because the next recession, asset bubble, or geopolitical crisis will not come with the fiscal buffer the U.S. has relied on for the last 80 years.
The disease is debt. And there are only two possible treatments: cut spending to levels that no Congress will ever vote for, or grow GDP fast enough to make the debt manageable.
The Treatment
This is where AI enters the picture — not as a Silicon Valley narrative about chatbots and stock prices, but as a potential structural shift in the American economy large enough to alter the fiscal trajectory.
Goldman Sachs estimates that widespread AI adoption could increase U.S. productivity growth by 1.5 percentage points annually over a 10-year period. The Penn Wharton Budget Model projects AI will raise GDP by 1.5% by 2035 and nearly 3% by 2055 — and in preliminary analysis, estimates that AI-driven productivity gains could reduce federal deficits by $400 billion over the 2026–2035 budget window alone. The St. Louis Fed is already tracking AI-related investment as a measurable contributor to GDP growth, with information processing equipment, software, R&D, and datacenter construction all contributing above their long-run averages.
IDC projects that AI will drive 3.5% of global GDP by 2030, with a cumulative economic impact of $19.9 trillion through the end of the decade. Every new dollar spent on AI solutions is projected to generate $4.60 in indirect and induced economic effects.
These are not speculative projections from AI evangelists. These are estimates from the institutions that central banks, pension funds, and sovereign wealth funds rely on for capital allocation decisions. And the consistent finding across all of them is the same: AI's economic impact depends entirely on infrastructure. Models don't run on ambition. They run on compute. Compute runs on power. Power runs on construction. Construction runs on the physical infrastructure — fiber, substations, cooling systems, water, gas — that companies like ours build every day.
The datacenters that Sanders and Ocasio-Cortez want to pause are the physical embodiment of the only GDP growth catalyst large enough to matter at the scale of a $39 trillion problem.
The Infrastructure Nobody Sees
Here is what the moratorium conversation consistently gets wrong: it treats datacenters as isolated buildings that benefit only Big Tech. In reality, datacenters are the demand signal for an entire industrial ecosystem.
Every hyperscale datacenter requires an average of 135 route miles of new fiber interconnection. The Fiber Broadband Association reports that U.S. fiber route miles need to nearly double — from 95,000 to 187,000 — by 2029. Datacenter bandwidth purchases increased 330% between 2020 and 2024. That fiber doesn't just serve the datacenter. It serves every community, every business, and every school along the route.
Every datacenter requires power infrastructure that the grid was never designed to support. Substations. Transmission lines. Natural gas generation. The U.S. power grid's infrastructure averages over 40 years old. Over 70% of power transformers are 25 years or older. Datacenters are forcing the grid upgrades that should have happened decades ago — and the construction industry is the one doing the work.
January 2026 set a record: $25.2 billion in datacenter construction starts in a single month. Trailing twelve-month starts totaled $103.7 billion. Datacenter employment is projected to reach 650,000 jobs this year. A single campus can require 4,000 construction workers, with an economic output multiplier of roughly 3.2x the initial investment. In Wisconsin alone, three companies have generated over $1 billion in equipment sales to datacenters.
These are not tech jobs. These are electricians, linemen, fiber splicers, directional drill operators, pipe layers, heavy equipment operators, and professional engineers. A moratorium doesn't pause Nvidia's stock price. It pauses their paychecks.
The Geopolitical Cliff
While Congress considers whether America is building too fast, China is not deliberating.
China is projected to hold 400 gigawatts of spare power capacity by 2030 — triple the expected power demand of the entire global datacenter fleet. Beijing's “Eastern Data, Western Computing” initiative is actively relocating energy-intensive datacenters to resource-rich western regions powered by surplus renewable energy. Chinese facilities pay less than half of what American datacenters pay for electricity, and projects move from planning to operations in months rather than years.
Nvidia CEO Jensen Huang has said publicly that building a datacenter in the United States takes roughly three years from groundbreaking to operation. His assessment of China's construction speed: they can build a hospital in a weekend.
The U.S. currently leads. It hosts nearly 5,500 datacenters — ten times China's count — and controls roughly three-quarters of global computing power. Capital expenditures by Microsoft, Amazon, Alphabet, and Meta are projected at $650 billion this year. Moody's projects $3 trillion in global datacenter spending over the next five years.
That capital goes wherever it can be deployed. If the United States signals that it is reconsidering whether to build, the capital migrates to the Middle East, Southeast Asia, and anywhere else the permitting environment doesn't require an act of Congress.
A moratorium doesn't pause the global AI race. It cedes America's position in it — and with it, the GDP growth that is the only realistic path to fiscal sustainability.
What the Debt Math Actually Requires
The CBO's baseline projections assume roughly 2% real GDP growth. At that rate, debt-to-GDP ratios continue to deteriorate indefinitely. The interest rate on federal debt exceeds the growth rate by 2031, and the gap widens every year after.
The only variable that changes the trajectory is growth. Not 2%. Not 2.5%. The kind of structural productivity acceleration that Goldman, Wharton, and the Fed are all tracking as a possibility — but only if the infrastructure gets built.
AI-driven productivity gains are already contributing above long-run averages to GDP through investment in software, R&D, information processing equipment, and datacenter construction. The question is not whether AI will have an economic impact. The question is whether the impact will be large enough, fast enough, to bend the debt curve before the debt curve breaks the economy.
Penn Wharton's Kent Smetters has cautioned against expecting AI to solve everything: “There's this belief among policymakers that in this new era of AI, we don't have to be fiscally responsible because AI is going to solve everything.” He's right that AI alone is not sufficient. But the inverse is also true — fiscal responsibility alone is not sufficient either, because the political system has demonstrated for 23 consecutive years of deficit spending that it will not cut its way to solvency.
If the only realistic path out is growth, and if AI is the most plausible catalyst for a productivity regime shift, then every policy that accelerates AI infrastructure deployment is a policy that improves America's odds of surviving its own debt. And every policy that slows it — including symbolic moratoriums introduced by legislators who have no intention of passing them — moves the timeline in the wrong direction.
What Would Actually Help
The real constraint on datacenter growth is not a lack of regulation. It is a lack of construction capacity.
Only 15% of applicants for datacenter construction roles meet minimum qualifications. The industry projects 340,000 unfilled positions by end of year. The skilled labor shortage across telecommunications, power, water, and gas construction predates the AI boom by decades — and it is the actual bottleneck.
If Congress wants to do something productive about the datacenter buildout, the answer is workforce development, not moratoriums. Accelerate apprenticeship programs. Streamline permitting. Fund the training pipelines that produce licensed electricians, certified fiber technicians, and qualified heavy equipment operators. Invest in the grid infrastructure that makes datacenter growth sustainable rather than parasitic.
Amazon has committed $700 million to skills training targeting 15,000 workers. Microsoft is partnering with community colleges to certify 8,500 technicians annually. Google has invested $50 million in apprenticeships across 12 metros. The private sector is spending billions to close the gap. The last thing the workforce needs is a congressional moratorium telling them the jobs are on hold while Congress studies the problem.
The Bottom Line
America's debt is not a talking point. It is a $39 trillion structural crisis compounding at $8 billion per day, with interest payments that already exceed defense spending and a debt-to-GDP trajectory that CBO projects will reach levels never before seen in the nation's history.
AI may not be a guaranteed cure. But it is the most credible candidate for the kind of productivity revolution that could generate the GDP growth necessary to stabilize the fiscal position. And that revolution depends entirely on physical infrastructure — fiber, power, water, gas, and the datacenters that consume all four.
A moratorium on datacenters is not a pause. It is a concession — to China, to the debt, and to a future in which America's greatest strategic advantage was abandoned because two legislators thought the country was building too fast.
We have never, in the history of this nation, grown our way out of a crisis by building less.
Legion Engineering, LLC is a Texas-licensed professional engineering firm (PE License #23937) providing design, permitting, construction, and project management services across telecommunications, power, water, and natural gas infrastructure in 25 states.